type originated by Budyko (30) and Sellers (51). - (,L977); __ and J. Weertman, ibid. 83, 4123The same experimental strategy has been used (1978).with general circulation models capable of simu- 59. N. Calder, Nature (London) 252, 216 (1974).lating the atmospheric response in three dimen- 60. J. Smagorinsky, in Meteorological Challenges:sions (52). So far, only the simpler, zonal models A History, D. P. McIntyre, Ed. (Informationhave been used in an alternative strategy de- Canada, Ottawa, 1972), p. 29.signed to simulate the climatic response to orbit- 61. For most insolation curves the error is less thanal variations. These experiments are discussed I percent. The exceptions are in high latitudes,in the following section. where there are seasons of continual darkness or

47. S. H. Schneider and R. E. Dickinson, Rev,- light.Geophys. Space Phys. 78, 447 (1974); R. G. Bar- 62. Weertman (56) attributes the difference to twory, Palaeogeogr. Palaeoclimatol. Palaeoecol. effects. First, even if the accumulation rate on a17, 123 (1975). growing ice sheet equals the ablation rate on a

48. G. R. North, J. Atmos. Sci. 32, 2033 (1975). shrinking ice sheet, the rate of change in ice vol-49. R. S. Lindzen and B. Farrell, ibid. 34, 1487 ume will be greater on the shrinking sheet than

(1977). on the growing sheet because the shrinking50. R. E. Newell, Quat. Res. (N. Y.) 4, 117 (1974); sheet is more spread out. Second, the accumula-

B. Saltzman and A. D. Vernekar, ibid. 5, 307 tion rate is expected to be smaller than the abla-(1975). tion rate.

51. W. D. Selers,J. Appl. Meteorol. 8, 392(1969). 63. J. Weertman, J. Glaciol. 13, 3 (1974); T.52. J. Williams, R. G. Barry, W. M. Washington, Hughes, G. H. Denton, M. G. Grosswald, Na-

ibid. 13, 305 (1974); W. L. Gates,J. Atmos. Sci. ture (London) 266, 596 (1977); R. H. Thomas33, 1844 (1976); Science 191, 1138 (1976); S. and C. R. Bentley, Quat. Res. (N.Y.) 10, 150Manabe and D. G. Hahn, J. Geophys. Res. 82, (1978); J. T. Andrews, Arct. Alp. Res. 5, 1853889 (1977). (1973).

53. In the simplest models, correlations are sought 64. However, there is evidence that ice sheets ofwith irradiation curves for a particular latitude modest size, such as those represented by cer-and season. Many workers, for example, have tain substages of isotopic stages 5 and 7, canfollowed (3) and chosen summer half-year irra- grow rapidly.diation at 65N [C. Emiliani, J. Geol. 74,109 65. Expressed in terms of half-response times, the(1966); W. F. Ruddiman and A. McIntyre, Sci- parameters Tm, Tc, and Tw are 11,800, 29,500,ence 204, 173 (1979)]. Others have chosen the and 7,300 years, respectively.same season at 45N (34, 35) or 50°N (11), or 66. W. S. Broecker, Science 189, 460 (1975); J. M.winter half-year irradiation at 55°N [G. J. Kukla, Mitchell, Jr., "Carbon dioxide and future cli-Curr. Anthropol. 9, 37 (1968)]. Another ap- mate," EDS (Environ. Data Serv.) (Marchproach has been to use linear combinations of 1977), p. 3.various individual season-latitude curves, in- 67. W. Q. Chin and V. Yevjevich, Colo. State Univ.cluding the following curves: summer irradiation (Fort Collins) Hydrol. Pap. 65 (1974).at 65N and 65S [W. Wundt, Quartdr Bibl. 68. Although the character of climatic records is(Bonn) 10/11, 15 (1961)], the hemispheric aver- fairly constant over the past 600,000 years (7,age summer irradiation and the irradiation for 38), older Pleistocene records are quite dif-summer at 70°N (54), and the annual irradiation ferent. For example, isotopic records in the in-at 65N combined with that for winter between terval from 600,000 to 2 million years ago have25N and 75N [G. J. Kukla, Boreas 1, 63 much reduced amplitudes and lack the 100,000-(1972)]. year cycle (41). Because the nature of orbital

54. R. G. Johnson and B. T. McClure, Quat. Res. variation is thought to have remained constant(N. Y.) 6, 325 (1976). over the past 2 million years, we conclude that

55. The need for a differential model has been recog- to understand these long climatic records, it maynized since the time of Milankovitch (3). Al- be necessary to use models whose parametersthough an extensive literature exists on what is vary with time (69).called the retardation problem [F. E. Zeuner, 69. V. Y. Sergin, Oregon State Univ. Clim. Res.The Pleistocene Period (Hutchinson, London, Inst. Rep. 2 (1978).1959), pp. 200-202; R. M. Fairbridge, Ann. N. Y. 70. J. M. Mitchell, Jr. (45), for example, doubts theAcad. Sci. 95, 542 (1961); W. S. Broecker (35)], existence of climatic resonance. But Sergin (69)only recently has a time-dependent model of the finds that there is a tendency to oscillate at peri-astronomical theory been formulated in terms of ods on the order of several tens of thousands ofdifferential equations describing the system re- years in a complex model of the climate system.sponse. See also E. Killrn, C. Crafoord, M. Ghil, J. At-.

56. J. Weertman, J. Glaciol. 6, 145 (1964). mos. Sci. 66, 2292 (1979).57. Weertman (56) used a physical model to esti- 71. System functions of this kind are discussed in E.

mate that the characteristic growth times of Eriksson, Meteorol. Monogr. 8, 68 (1968).large ice sheets are on the order of 15,000 to 72. E. N. Lorenz, ibid. 8, 1 (1968); Tellus 3, 28930,000 years. See also J. T. Andrews and M. A. (1969).W. Mahaffy, Quat. Res. (N. Y.) 6, 167 (1976). 73. Other possible deterministic sources of variabili-

58. G. E. Birchield, J. Geophys. Res. 82, 4909 ty will also have to be explored (45). However,since orbital variation is now the only forcing

DNA Gyrase and theSupercoiling of DNA

Nicholas R. Cozzarelli

Virtually all duplex DNA exists natu-rally in a negatively supertwisted form.This supertwisting is an important facetof the processes of DNA replication,transcription, and recombination. Just 4

SCIENCE, VOL. 207, 29 FEBRUARY 1980

years ago, a bacterial enzyme that in-troduces negative supertwists into DNAwas discovered by Gellert et al., whochristened it DNA gyrase (1). This ar-ticle describes how the rapidly advanc-

function that can be specified exactly, it is bestto exhaust the explanatory power of the astro-nomical theory first.

74. Building stochastic models to understand thevariations unexplained by deterministic modelsis a relatively undeveloped field for the range offrequencies under consideration in this article(12). A model proposed by Hasselmann (43) gen-erates a red-noise oceanic response to white-noise atmospheric forcing. Almost-intransitivemodels (72) provide another potentially fruitfulavenue of research.

75. Two input parameters, corresponding to thechoice of latitude and season (or a and O), areassigned to models that use a single input curve.If the scale of the input is important, then a thirdparameter is assigned. All other parameters arecounted in cf. For the linear model there is onetime constant only; for the nonlinear modelthere are warming and cooling time constants;for Calder's model there is a ratio of warmingand cooling response rates and a critical value ofthe input; and for Weertman's model there areaccumulation and ablation rates, basal shearstress, slope of the snow line, and a critical val-ue of the input.

76. J. Iversen, Dan. Geol. Unders. 5 (1973); J. C.Bernabo and T. Webb. Quat. Res. (N. Y.) 8, 64(1977); H. E. Wright, Jr., Annu. Rev. EarthPlanet. Sci. 5, 123 (1977).

77. N. G. Pisias, Geol. Soc. Am. Mem. 145 (1976),p. 375, table 2.

78. . J. Mesolella, R. K. Matthews, W. S. Broeck-er, D. L. Thurber, J. Geol. 77, 250 (1969).

79. N. J. Shackleton, Proc. R. Soc. London. Ser. B174, 135 (1969).

80. All spectra have been calculated by using stan-dard autocorrelation procedures identical tothose described in (5). Let N be the number ofsample points, m be the number of lags, At bethe sampling interval in K years, and C be theconstant of a first-difference filter. For the modelinput and output spectra in Fig. 9, A and B,N =501, m = 199, and At = 2K. For the eccentricityspectrum in Fig. 9A, N = 551, m = 199, and At= 2K. In Fig. 9C,N = 157, m = 50, A = 3K,and C = 0.998 for the prewhitened spectrum. InFig. 9D, N = 122, m = 61, and At = 6K for theunprewhitened spectrum and N = 122, m = 47,A = 6K, and C = 0.998 for the prewhitenedspectrum. In Fig. 9E, N = 367, m = 133, and At= 3K. Confidence intervals and bandwidthshave been calculated by procedures given in G.M. Jenkins and D. G. Watts [Spectral Analysisand its Applications (Holden-Day, San Fran-cisco, 1968)].

81. Supported by NSF grants OCD7S-14934 andATM77-07755 to Brown University. K. Bryan,I. M. Held, J. M. Mitchell, Jr., T. C. Moore, Jr.,W. F. Ruddiman, T. Webb III, and J. Weertmancritically read an earlier version of this articleand made valuable suggestions for improving it.We thank R. M. Mellor and T. A. Peters for pre-paring the typescript. This paper will be present-ed as a Richard Foster Flint lecture at Yale Uni-versity, New Haven, Conn., on 4 March 1980.

ing studies of gyrase have traversed a

spectrum of topics of contemporary in-terest including the mechanism of super-coiling, the energetics of macromolecu-lar movement, the conversion of DNAinto complex topological forms, the site-specific binding of enzymes to DNA, thereversible association of enzyme sub-units, and the mechanism of inhibitors ofDNA synthesis. It is a propitious timefor a review of gyrase, since critical fea-tures of the enzyme can now be ex-

plained satisfactorily by the recently pro-posed mechanism termed sign inversion(2). This article focuses on the enzymefrom Escherichia coli about which mostis known (3).

The author is a professor of biochemistry at theUniversity of Chicago, 920 East 58 Street, Chicago,Illinois 60637.

0036-8075/80/0229-0953$01.75/0 Copyright G 1980 AAAS 953

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Sep

tem

ber

13, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Background

The two polynucleotide strands ofduplex DNA wind around each other in aright-handed helix containing about onehelical turn every ten base pairs. Super-coiling, the coiling of the helix axis itself,was discovered by Vinograd et al. in astudy of closed, circular DNA (4). Thesense of supercoiling can, in principle,be either the same or opposite to that ofthe helical twists which, by convention,are designated as positive. Remarkably,all supercoiled DNA isolated from natu-ral sources is twisted opposite to the di-rection of the double helix, with aboutone negative supercoil per 15 double he-lical twists (5). There is a fundamental

The c protein relaxes positively super-coiled DNA very poorly (8, 10). Thismay seem a trivial deficiency since DNAin the cell is negatively supercoiled. It is,in fact, important in the cell to maintainnegative supertwists and to relieve posi-tive twisting stress that is generated byprocesses, such as DNA replication, thatunwind the double helix. Champoux andDulbecco discovered a topoisomerase inmouse cells, which relaxes positivelyand negatively supercoiled DNA withequal facility (I1). Similar topoisomer-ases have since been identified in manyeukaryotic organisms, and enzymes withthe restricted relaxation pattern of w arelimited to bacteria where their functionremains an enigma (12).

Summary. Negative supercoiling of bacterial DNA by DNA gyrase influences allmetabolic processes involving DNA and is essential for replication. Gyrase supercoilsDNA by a mechanism called sign inversion, whereby a positive supercoil is directlyinverted to a negative one by passing a DNA segment through a transient double-strand break. Reversal of this scheme relaxes DNA, and this mechanism also ac-counts for the ability of gyrase to catenate and uncatenate DNA rings. Each round ofsupercoiling is driven by a conformational change induced by adenosine triphosphate(ATP) binding; ATP hydrolysis permits fresh cycles. The inhibition of gyrase by twoclasses of antimicrobials reflects its composition from two reversibly associated sub-units. The A subunit is particularly associated with the concerted breakage-and-re-joining of DNA and the B subunit mediates energy transduction. Gyrase is a prototypefor a growing class of prokaryotic and eukaryotic topoisomerases that interconvertcomplex forms by way of transient double-strand breaks.

topological constraint on any DNAmolecule, such as closed, circular DNA,whose polynucleotide strands cannot ro-

tate freely about the helix axis: the alge-braic sum of the number of supercoilsand the number of double helical turns is

a constant, the linking number (6). Thus,unwinding of the double helix must becompensated by an equivalent windingup-that is, removal-of negative super-coils. Since the supercoiled configura-tion is strained, the removal of super-coils is highly energetically favorable.Negative supercoiling thereby promotesthe binding of any protein that locallydenatures (unwinds) DNA; this occurs

with a number of enzymes of DNA me-

tabolism (7).Wang initiated the enzymology of su-

percoiling with the discovery ofE. coliprotein (8). Requiring no nucleotide co-

factor, X removes (that is, relaxes) mostof the negative supercoils in DNA. Sincethe linking number is changed in this re-

action, a transient break in the DNA isrequired. The co protein is the archetypeof the important class of enzymes knownas topoisomerases, which catalyze theinterconversion of DNA molecules thatdiffer only in linking number (9). Thus E.coli X protein has been designated EcoDNA topoisomerase I (topo I).

DNA gyrase (topo II) is the onlyknown topoisomerase that catalyzes su-

percoiling of DNA. It was discovered ina remarkably logical way. Purified intprotein from bacteriophage X requiresnegatively supercoiled DNA to promoterecombination in vitro (13). However,relaxed DNA sufficed with a crude prep-aration of int if adenosine triphosphate(ATP) was provided (14). The search foran ATP-dependent supercoiling activitycontaminating the int preparation was re-warded by the discovery of DNA gyrase(1, 14). The importance of DNA gyrasein both DNA replication and in the in-troduction of negative supercoils was

soon established. Gellert et al. foundthat novobiocin and coumermycin A1,two closely related inhibitors of nucleicacid synthesis in vivo, were inhibitors ofDNA gyrase in vitro (15). Moreover,coumermycin A1 inhibited the super-coiling of circular X DNA in vivo. Bothof these effects were abolished by a mu-

tation in the cou gene, which controls thecellular level of sensitivity to these drugs(16).My involvement with DNA gyrase re-

sulted from a study of the mechanism ofa different pair of inhibitors of DNA rep-lication-nalidixic acid and the more po-tent analog, oxolinic acid. The target

protein for these drugs in E. coli, thenalA gene product, was purified (17) bymeans of a complementation assay forkX174 DNA replication (18, 19). Thediscovery that gyrase from wild-type butnot nalA mutant cells was inhibited bynalidixic acid (17, 20) established the sur-prising link between the nalA protein andDNA gyrase.

Structure of Gyrase

The physical basis for the segregationof gyrase inhibitors into two classes isthe construction of the enzyme from twosubunits, A and B, which are the targetproteins for nalidixic acid and novobio-cin, respectively (21, 22). Micrococcusluteus gyrase subunits a and /8 are func-tionally analogous to A and B, respec-tively (23, 24). Establishment of the com-position of gyrase was hampered by lowpurification yields until independentstudies on the very different micro-organisms E. coli (25) and M. luteus (23)revealed that only a small proportion ofthe subunits are associated in extracts.Mixing of the subunits under appropriateconditions efficiently reconstitutes gy-rase activity and provides a convenientassay that has been used in the prepara-tion of both subunits in physically ho-mogeneous form (25, 26). There is per-haps ten times more subunit A than sub-unit B in cells (25); as discussed below,subunit A is also part of another topo-isomerase. Subunit A is a homodimer of105,000-dalton nalA protomers and sub-unit B consists of 95,000-dalton cou pro-tomers (25). The small amount of consti-tuted enzyme in E. coli cell extracts hasbeen purified as such and contains justthe same two polypeptides (25, 27).

Convincing evidence for the structuralgene assignments led to the renaming(28) of nalA and cou as gyrA and gyrB.First, by two-dimensional electropho-resis under denaturing conditions, sub-unit A and radioactive gyrA gene prod-uct are identical (21), as are subunit Band authentic gyrB gene product (28).Second, subunit A from a gyrA mutantconferring nalidixic acid resistance andsubunit B from a novobiocin-resistantgyrB mutant each reconstitute a gyrasewith the expected drug resistance (25).Third, subunit A purified from a gyrAtemperature-sensitive mutant is ther-molabile (29, 30). Fourth, attachment ofa reactive ATP derivative to subunit B ofgyrase is specifically inhibited by no-vobiocin (27). .

Although the protomer structure of gy-rase is not rigorously established, it al-most surely is of the form a2,32. There isgood evidence that gyrase contains two

gyrA protomers, each of which has an

active site (21, 31), and that gyrase con-

tains equivalent amounts of gyrA andgyrB protomers (27, 31). Equal molaramounts of subunits A and B reconsti-tute gyrase activity, and subunit A, atleast, is a homodimer (17, 25). Finally,the best estimate for the molecularweight of the holoenzyme is 400,000, thevalue expected for an a2f32 structure (27,32).

Activities ofDNA Gyrase

Gyrase remains bound to its DNA sub-strate, supercoiling it processively (33)and catalytically (25, 27). One moleculeintroduces about 100 supertwists per

minute at 30°C. Under standard reactionconditions, the supercoiling of Col ElDNA reaches a limit at one and a halftimes the density of Col El DNA super-

coiled in vivo (1). Both positively super-coiled and relaxed DNA can be negative-ly supercoiled (21, 24).The supercoiling reaction has two

components. Since the linking number ofthe DNA is changed, there must be a

transient break of one or both strands(5). In addition to this concerted break-age-and-reunion activity (34), gyrasemust have a coupled energy transductioncomponent since supercoiling is an en-

dergonic reaction. Some of the reactionsof gyrase described below isolate one ofthese components uncomplicated by theother. All the reactions have been impor-tant in understanding the enzyme.

Adenosine triphosphatase (ATPase).Like many enzymes of DNA metabolism(35), gyrase hydrolyzes ATP to adeno-sine diphosphate (ADP) and Pi in thepresence of DNA (19, 26, 27, 36). Thereaction is highly specific for ATP amongnaturally occurring nucleotides, as is su-

percoiling. Duplex DNA is about a tentimes better effector than single-strandedDNA' in accord with the preferentialbinding of gyrase to duplex DNA. Asimple interpretation of the effector ac-

tivity of DNA is that its binding stabiliz-es an enzyme conformation that hasATPase activity. The ATP hydrolysis isnot necessarily coupled to supertwistformation and continues after maximumcoiling of a DNA effector. The potent in-hibition of ATPase by novobiocin butnot by oxolinic acid makes it likely thatthe energy transduction component is as-

sociated intimately with the B subunit ofgyrase (37).

Relaxation. Gyrase spontaneously re-

laxes negative supercoils in the absenceof ATP (17, 20) but cannot relax positivesupercoils (23, 24). This asymmetry mayreflect the importance of handedness in

the sign inversion mechanism for gyrase.The rate of relaxation is an order of mag-nitude lower than the rate of super-

twisting (25) and therefore under-estimates the breakage-and-reunion ca-

pacity of gyrase. A plausible explanationis that breakage-and-reunion is usuallycoupled to supertwist formation and thusis stimulated by ATP. Relaxation is in-hibited by nalidixic acid but not by no-

vobiocin, just the opposite to resultswith ATPase. Therefore, subunit A isparticularly important for the breakage-and-reunion component of DNA gyrase.Cleavage. Nalidixic acid inhibits gy-

rase by interfering with the breakage-and-reunion component of supercoiling.Addition of a protein denaturant to a

nalidixic acid-treated gyrase reaction re-

sults in breakage of both polynucleotidestrands and concomitant attachment ofenzyme to the cleaved DNA. Gyrase dis-rupts the DNA strands at points stag-gered by four base pairs, creating terminithat each possess a free 3' hydroxylgroup and a 5' extension attached cova-

lently to a gyrA protomer (31, 38). Thesetermini therefore provide a template-pri-mer for DNA polymerases but are com-

pletely resistant to 5' terminal labelingwith bacteriophage T4 polynucleotide ki-nase (38). Gyrase thus contains two gyrAprotomers each of which makes up a crit-ical portion of the breakage-and-reunionactive sites. The covalent interaction be-tween gyrase and DNA can be under-stood in terms of Wang's model (8) forthe way in which topoisomerases, unlikeclassical DNA ligases (39), reseal tran-sient breaks in DNA without infusion ofenergy from a nucleotide cofactor. In-stead of irreversible hydrolysis of a

phosphodiester bond, it is proposed thatstrand breakage occurs in a transfer re-

action in which the enzyme itself be-comes attached covalently to the brokenend. Resealing in the reverse transfer re-

action releases free enzyme. Thisscheme is supported by the general ob-servation that various conditions un-

couple breakage from reunion by topo-isomerases and result in covalent joiningof enzyme to DNA (12). The discoveryof breakage of both DNA strands by gy-rase (17, 20) set it apart from other topo-isomerases that nick only one strand,and implies that the mechanism of gyraseis fundamentally different.A further distinction of gyrase from

nearly all other topoisomerases is thatcleavage is highly site-specific (17, 20,38). With less than one gyrase. moleculeper DNA substrate of about 4 to 5 x 106daltons, most often cleavage is at one or

a few sites. The balance occurs withwidely varying frequency at a number ofother positions so that potential cleavage

sites occur, on the average, about onceper 100 base pairs (38). At each site, thecuts are specific at the nucleotide level.Cleavage can be induced by reaction per-turbants other than nalidixic or oxolinicacid and is observed at low frequency inthe unperturbed reaction (21, 24, 40). Allthese different conditions result in cleav-age at the same sites (24). These sites arealso the ones cut by M. luteus and Ba-cillus subtilis gyrases, by a hybrid gyraseconstructed from M. luteus and E. colisubunits, and by E. coli topo II', an en-zyme related to gyrase (24, 41). This con-servation suggests that site specificityplays a critical role in supercoiling (42).The cohesive end sequences of six gy-

rase cleavage sites have been determined(38, 43). The dinucleotide TG (T, thy-mine; G, guanine), straddling the gyrasecut on one of the DNA strands, providedthe only common bases within a 250base-pair region surrounding the sites.Analysis of other sites showed that cut-ting between a TG doublet is common tonearly all gyrase cleavages. Other basescommon to some of the sequenced sitesare clustered nonrandomly around theTG doublet and may be variable com-ponents of the cleavage sequence, but nosimple set of related sequences has beenidentified that unfailingly predicts sitesof cleavage. This pattern of site specifici-ty in the absence of a unique determiningsequence is emerging as a common fea-ture in nucleic acid-protein recognition.

In contrast to the conservation of sitesof cleavage, the degree of cleavage atany site varies with reaction conditions.ATP specifically incites a redistributionof cleavage pattern, stimulating cleavageat some sites while reducing it at others(36). This phenomenon might have sug-gested movement of gyrase betweencleavage sites, either by solution or bytranslocation along the DNA. Suchmovement was ruled out (33), however,since cleavage site redistribution wasunaffected by competitor DNA and sinceATP still stimulated cleavage at individ-ual sites isolated in small (176 to 509 basepairs) DNA fragments. Furthermore,ATP did not alter the binding of gyrase tosmall DNA fragments although it stimu-lated cleavage by as much as an order ofmagnitude. Gyrase is thus envisioned asremaining stationarily bound to DNA atdiscrete locations; ATP effects a con-formational change in the enzyme thatalters, in a fashion dependent on thelocal DNA sequence, the proportion ofbound gyrase that cleaves its substrate.Binding to DNA. The binding of gy-

rase to radioactively labeled DNA is eas-ily measured by the resulting retention ofthe label by nitrocellulose filters (21, 44).This complex of gyrase with DNA, un-

like that formed by cleavage, is stabi-lized by noncovalent bonds that are dis-rupted by high ionic strength or proteindenaturants. It is nonetheless a stablecomplex, since gyrase dissociates fromlinear Col El DNA with a half-life ofdays at 23°C. Neither gyrase subunit byitself binds detectably to DNA.Gyrase binds to DNA site specifically.

The sites at which gyrase cleaves DNAare generally a faithful reflection ofwhere it binds, since almost 90 percentof the enzyme bound noncovalently canalso cleave the DNA (33). Site-specificbinding implies that gyrase also acts invivo only at certain sites. The limitedcleavage following oxolinic acid additionin vivo is consistent with this conclusionand may provide an opportunity to mapthe sites of gyrase action in the cell (45).Given that gyrase acts at discretechromosomal locations, there may be re-gional differences in processes that gy-rase facilitates, depending on the loca-tion and strength of both gyrase bindingsites and the barriers that delineate su-percoiling domains in the cell (46).Gyrase binds less tightly to DNA that

is negatively supercoiled than to relaxedDNA (44), and thus the stable complexwith relaxed DNA dissociates rapidlywhen addition of ATP provokes super-coiling (33). This feedback loop provides

01.

2.

a mechanism for turnover of an other-wise stably bound enzyme and for limit-ing the final supertwist density of DNA.

Liu and Wang (23, 47) have proposedthat when DNA gyrase binds to DNA theDNA becomes wrapped around the en-zyme in a positive coil. This was basedinitially on the increase in linking num-ber imparted by bound DNA gyrase tonicked circular DNA converted to thecovalently closed form by DNA ligase.Supporting evidence for the coiling ofDNA around gyrase is provided by thesimilarity in the pattern of protectionfrom nucleases afforded by gyrase to thatobtained with nucleosomes where wrap-ping has been established.Catenation and uncatenation. The

most recently discovered of the gyrasereactions are catenation-the inter-locking of duplex DNA rings-and theresolution of catenanes into componentcircles (40). Catenanes are physiological-ly important structures that are the con-stant companion of circular DNA mole-cules and are sometimes the pre-dominant cellular species; they have fre-quently been suggested as intermediatesin recombination or replication (48, 49).As in supercoiling and relaxation, thelinking number is changed in these reac-tions but at the level of quaternary ratherthan tertiary structure. Catenation and

NovATP

gy rase

DNA + gyrase -k DNA.gyrase

3. (ooOx

4.gyra

Gyrase

o>0 - >

ase N aDodSO4

uncatenation provide the first direct evi-dence that gyrase makes a double-strandbreak, passes the DNA through thebreak, and reseals the break in an effi-cient, concerted reaction. The alterna-tive of progressive invasion of one ringby another seems unlikely in view of thedouble-strand breaks resulting fromcleavage. The analogous reaction-re-moval of knots from duplex DNA cir-cles, catalyzed by several topoisomer-ases, including gyrase-is also inter-pretable in terms of a transient double-strand break (50, 51).

Catenation and uncatenation are themost fastidious of the gyrase reactions,and this helps explain why they weremissed in earlier work. There is a sharpionic strength optimum, and ATP, Mg2+,and spermidine are required. However,the reactions are surprisingly effectivesince most of the substrate can be cate-nated or decatenated. Negatively super-coiled DNA is more efficiently inter-locked than nicked, circular DNA. Ho-mology between donor and acceptorDNA is not required since catenanes areformed between unrelated DNA mole-cules. The number of rings linked andunlinked progressively changes withtime, and molecules containing as manyas thousands of rings have been ob-served.Summary of activities. The activities

of DNA gyrase and their sensitivity toI n h i bitor s the two families of drugs are summarized

fobiocin Oxo in Fig. 1. Only those reactions requiringY es Yes ATP (supercoiling, ATPase, catenation,

and decatenation) are inhibited by no-vobiocin, and the reactions that requirebreakage-and-reunion (supercoiling, re-

N o N o laxation, catenation, and decatenation)are selectively inhibited by oxolinic acid.DNA is either a substrate or an effectorfor each reaction. Reconstitution of all

N O Y es the activities requires an equal amount ofsubunits A and B. The need for both sub-units for binding to DNA is sufficient to

No N o explain this requirement for all the otherreactions.

Duplex DNA5. ATP > ADP + P

gy r ase

6. 2 ATP, gyrase

Yes N o Energy Coupling and the Mechanism ofAction of Novobiocin

Yes Yes

Fig. 1. The activities of DNA gyrase. The reactions depicted are supercoiling (reaction 1),binding (reaction 2), relaxation (reaction 3), cleavage (reaction 4), ATPase (reaction 5), andcatenation and uncatenation (reaction 6). The DNA substrate is shown as relaxed, duplex,circular DNA for reaction 1, negatively supercoiled DNA for reactions 3 and 6, and as linearDNA for reaction 4. Oxolinic acid is abbreviated Oxo and sodium dodecyl sulfate as Na-DodSO4. Nalidixic acid has the same inhibition spectrum as oxolinic acid, and coumermycin A1has the same spectrum as novobiocin. Second-order effects, such as the low stimulatory activityof single-stranded DNA for ATPase and the inhibition of ATPase by oxolinic acid under certainconditions (26), are not presented.

956

The key feature of the mechanism ofenergy coupling by gyrase was discov-ered serendipitously (36). A nonhydro-lyzable analog of ATP, adeny 1-5'-yl-imi-dodiphosphate [App(NH)p], caused ashift in cleavage pattern similar to thatevoked by ATP. This was surprising be-cause hydrolysis of ATP was known tobe required for other gyrase reactions,such as supercoiling (36). However,DNA gyrase can turn over in the super-

SCIENCE, VOL. 207

coiling reaction (25, 27) but not in thecleavage reaction, which requires en-zyme denaturation. The paradox wouldbe resolved, then, if the binding of ATP[or of App(NH)p] leads to a single roundof supercoiling, but ATP hydrolysis isneeded for enzyme turnover. This strongprediction was confirmed; with substratelevels of gyrase, App(NH)p led to super-coiling stoichiometric with the number ofenzyme molecules (36). The derivedstoichiometry of a little less than one su-percoil per gyrase molecule is a lowerlimit since it neglects inactive enzymeand the back reaction of relaxation.To explain these results, it is proposed

that binding of ATP or of App(NH)p re-sults in a conformational transition thatdrives a crucial part of the supercoilingcycle (36). For gyrase to return to itsoriginal conformation, that is, for turn-over, nucleotide desorption is required,and this is facilitated by hydrolysis ofATP to the less tightly bound ADP.This type of energy coupling was origi-

nally proposed to explain chemical ener-gy-fueled movement in other processes,such as muscle contraction, transferRNA translocation, active transport, andoxidative phosphorylation (52-55). Itshould be contrasted with the bioener-getic pattern first conceived by Lipmann(56) and Kalckar (57) in which a high-en-ergy intermediate is formed that consistsof part of a nucleoside triphosphate co-valently bonded to a substrate of the en-dergonic reaction, such as aminoacyladenosine monophosphate in proteinsynthesis. A monetary analogy clarifies amajor difference between the two bioen-ergetic schemes. The conformationalcoupling pathway is like using a creditcard since one can "buy now" (super-coil) but "pay later" (hydrolyze ATP);the Lipmann-Kalckar mechanism hasthe parsimony of pay-as-you-go. Recentexperiments suggest that the T4 topo-isomerase (51) and the E. coli recA pro-tein (58) also employ credit-card energet-ics. There are many other ATP-depen-dent enzymes in DNA metabolism (35),and these may also follow the gyrasemodel.Novobiocin selectively interferes with

the energy transduction component ofgyrase; reactions independent of ATPare immune (Fig. 1). The basis for thisspecificity was suggested by the findingthat novobiocin prevents reorientation ofthe cleavage pattern by ATP or byApp(NH)p (36). Thus, the antibioticmust act prior to ATP hydrolysis. Com-petitive inhibition with respect to ATP inboth the supercoiling and ATPase reac-tions showed that the sensitive step isATP binding (36). The same conclusion29 FEBRUARY 1980

Fig. 2. The sign inver-sion model for enzy-matic supercoiling of Stabiliz

-positivfDNA. The enzyme ndmaintains and acts atthe upper node; the(+) and (-) symbolsrefer to the sign of thenodes.

was reached from the prevention by no-vobiocin of labeling of subunit B by anATP analog (27). Novobiocin is an im-pressive inhibitor-the Ki (inhibitionconstant) of 10-8M is several orders ofmagnitude less than the Km (Michaelisconstant) for ATP of 0.3 mM (36). De-spite the selective and competitive effectof novobiocin on ATP binding, there isno obvious structural similarity betweenthem. Novobiocin may block ATP ac-cess without sharing binding sites, or theenzyme conformations competent forbinding ATP and novobiocin could be in-compatible.

Mechanism of Supercoiling

Early models for gyrase postulatedthat negative supercoils are introducedin two steps. An initial segregation ofpositively and negatively supercoiled re-gions of a DNA molecule is followed byselective relaxation of the positive super-coils, leaving just the negative supercoils(17, 23, 27). The critical observation con-cerning supercoil segregation is that thebinding of gyrase per se generates posi-tive supercoils in the absence of ATP;these are adjacent to the enzyme sincethey are immune to relaxing enzymes(21, 23, 47). Compensating negative su-percoils result in other regions of themolecule since the linking number mustbe unchanged. Addition of App(NH)pnot only causes limited negative super-coiling stoichiometric with the number ofgyrase molecules but also an equivalentreduction in protected positive super-coils (21). An explanation for this corre-lation is a causal link between removal ofthe positive supercoils of binding and in-troduction of negative supercoils.The second step in supercoiling, selec-

tive relaxation of positively supercoiledregions, has usually been described interms of the traditional topoisomerasemodel of nicking of one strand, rotationabout the helix axis to dissipate the posi-tive stress, and resealing of the nick (59).Recent work, however, has established acategorically different mechanism called" sign inversion" because the positivesupercoils are not relaxed but activelyinverted to negative supercoils (2). This

mechanism is illustrated in Fig. 2 andconsists of the following steps.

1) Gyrase binds to a DNA moleculesuch that the bound segments cross toform a right-handed node (the uppernode in Fig. 2). This stabilizes a positivesupercoil and induces a counterposingnegative supercoil represented by thelower (-) node.

2) Gyrase introduces a double-strandbreak in the DNA at the back of theright-handed node and passes the frontsegment through the break, inverting thehandedness and thus the sign of thenode.

3) The break is resealed on the frontside of the now left-handed node.Gyrase could then release one of the

two crossing segments of the negativenode, the negative supercoils would dis-tribute along the DNA, and the cyclecould begin again. The net result of thesteps illustrated in Fig. 2 is to reduce thelinking number of the DNA by two (60).Reversal of these steps, that is, startingwith a left-handed node and inverting itto a right-handed one, increases the link-ing number by two and is the proposedpath for relaxation of negative supercoilsby gyrase.The unique prediction of the model

has been verified; both supercoiling andrelaxation of DNA change the linkingnumber in steps of two (2). Since no odd-numbered changes in linking numberwere observed, the ends of the double-strand break in the intermediate do notrotate relative to one another and this en-sures that supercoils are not dissipated.Catenation, unknotting, and uncatena-tion substantiate the other characteristicfeature of the model, passing of DNAthrough a transient double-strand break(40, 51). These three related reactionsare, in turn, readily explicable by thesign-inversion model. Catenation, for ex-ample, results when the two crossingsegments of the node are contributed bydifferent DNA molecules.The following somewhat speculative

elaboration of the sign inversion mecha-nism incorporates the gyrase energytransduction scheme and features of gy-rase as revealed by studies ofDNA bind-ing and cleavage. Gyrase binds to DNAat specific DNA sequences, forming a

957

right-handed node. A positive coiling ofthe DNA around the enzyme is a plau-sible means for ensuring the properhandedness (23, 47), but the sequence ofthe crossing segments could provide thenecessary information (2). The sub-sequent binding of ATP by each of thegyrB protomers changes the con-formation of gyrase to one that stabilizesa left-handed node. The resultant strainis relieved by passing the front segmentof DNA through a double-strand breakmade in the back segment. Remarkably,this is accomplished while gyrase holdsboth ends of the break so that they can-not rotate. The gyrA protomers are at-tached by covalent bonds to each 5'-phosphoryl terminus of the break. Theenergy of the broken phosphodiesterbonds is conserved in this covalent com-plex and is used to reseal the break aftersign inversion. The cohesive ends of thebreak aid the proper alignment for re-sealing. Gyrase then catalyzes hydroly-sis of ATP, and the segment that waspassed through the break is released sothat resetting is possible. The desorptionof the loosely bound ADP causes the en-zyme to return to its initial con-formation, the more stable one in the ab-sence of a nucleotide effector. Gyrase re-mains bound at the same specific DNAsites (those at which it cleaves), actingprocessively through cycles of sign in-version until binding is sufficiently weak-ened by the increased negative super-twist density so that the enzyme is re-leased.

New Topoisomerase Activities

Related to Gyrase

Equimolar amounts of subunits A andB are needed to reconstitute both the su-pertwisting and relaxing activities of gy-rase. It was therefore puzzling that theratio of oxolinic acid-sensitive relaxa-tion activity to supertwisting activity de-clined (to a nonzero plateau value) withincreasing purity of the enzyme (17, 21,25). One explanation for this apparentparadox would be contamination of lesspure preparations with a relaxing en-zyme distinct from gyrase but sharingthe gyrA product that controls oxolinicacid sensitivity.Such an activity has been identified

and is designated topo II' (24). It is con-structed from subunit A and a 50,000-dalton protein called v. Since subunit Band v peptide maps are quite similar, vmay be a processed form of subunit B orderived from a transcript of part ofgyrB.Whatever the ontogeny of v, availableevidence implies that topo II' exists nor-

958

mally in the cell rather than being de-rived from subunit B during purification.Four separate purifications each detect-ed an order of magnitude more v thansubunit B (24), and similar levels of v arefound in the dissimilar E. coli K and Bstrains (61). Among E. coli topoisomer-ases, topo II' alone relaxes positive su-percoils as efficiently as it relaxes nega-tive supercoils. Topo II' has no negativesupercoiling activity and is not affectedby either ATP or novobiocin but closelyresembles gyrase in its other properties,including sensitivity to oxolinic acid, sta-bilization of positive supercoils, double-strand cleavage at the same DNA sites,and catenation and uncatenation ofDNArings. If the structural homology of v andsubunit B is confirmed, then it appearsthat subunit B is divided into functionaldomains. One domain represented by vis sufficient for binding subunit A and re-constitution of breakage and reunion.The other domain contains the ATPbinding site or allows its expression inenergy-requiring reactions. Its absencein topo II' somehow uncouples positivenode inversion from ATP binding and al-lows relaxation of positive supercoils.The discovery of a novel topoisomer-

ase in T4-infected cells explains somepuzzling results (50, 62). DNA synthesisdirected by almost all other bacterio-phages is strikingly reduced by inhibitionof topo II and topo II', but the reductionin T4-infected cells was only about five-fold and sometimes less (30, 63). How-ever, mutants in genes 39, 52, or 60, so-called DNA delay genes, are totally de-pendent on the host gyrase by the sametests (64). The products of these threegenes are the protomers of a T4 topo-isomerase that requires ATP not to su-pertwist DNA but to relax it, whether itis negatively or positively supercoiled(50, 62, 65). Very high levels of T4 topo-isomerase catenate supercoiled DNA in-tramolecularly (that is, knot it) in anATP-independent reaction; in the pres-ence of ATP the same enzyme can untiethe knot catalytically, even after relaxa-tion or nicking (51).

Gyrase and T4 topoisomerase canprobably perform the same function inT4 metabolism since each enzyme sparesthe requirement for the other. The lackof supertwisting by the phage enzyme istherefore intriguing. Perhaps what is cru-cial for the phage is an activity shared byboth topoisomerases, such as uncatena-tion or relief of positive twisting stress.Alternatively, there is substantial cir-cumstantial evidence that genes 39, 52,and 60 are required for proper initiationof T4 DNA replication (66). It has beensuggested (48) that the T4 enzyme might

be an origin-specific gyrase and thus su-pertwist such DNA.Whereas no eukaryotic enzyme has

yet been shown to supercoil DNA, gy-rase and the T4 enzyme are prototypesof the new class of topoisomerases thatinterconvert complex topological formsof DNA and make transient double-strand breaks. It was in an extract of aeukaryote (Xenopis laevis) that a cate-nation activity was actually first identi-fied several years ago by Attardi et al.(67); this activity has been shown to beATP dependent (68). A topoisomerasefrom early embryos of Drosophila cancatenate DNA (69) and untie knots pro-duced by T4 topoisomerase in circularDNA (51); both reactions require ATP.A vaccinia virus-induced enzyme andtrypanosome extracts can resolve the gi-ant catenated networks of trypanosomekinetoplast DNA (70).

Functions of Gyrase

The physiological role of gyrase hasbeen studied with the use of specificdrugs, structural gene mutations, and invitro DNA replication and recombina-tion systems. The isolation of gyrA andgyrB conditional lethal alleles demon-strated that DNA gyrase is an essentialenzyme forE. coli growth. The emergingconclusion is of a widespread impor-tance of the enzyme in DNA metabo-lism.There is now good evidence that gy-

rase is responsible for introducing atleast most of the negative supertwists invivo. Gyrase inhibitors block super-twisting of infecting phage X DNA (15,20, 71) and drastically reduce the super-helicity of the, folded E. coli chromosome(46, 47, 72). Intercalating agents removenegative supercoils and cause curing,that is, a preferential loss of plasmids(73). Sublethal doses of novobiocin cansimilarly cure (74, 75), and genetic im-pairment of gyrase enhances sensitivityto the intercalating agent, acriflavin (28).In fact, the discovery of gyrase and itseffect on supertwisting has provided themost compelling evidence that super-coils do exist in the cell and are not anartifact of protein loss or ionic changeson isolation of the DNA. Analogously tosupercoiling in eukaryotes (discussed be-low), the bacterial histone-like proteinscould make some contribution to super-twisting or help stabilize gyrase-inducedsupercoils (76).The striking physiological effect of in-

activation of gyrase is the complete inhi-bition of replicative DNA synthesis. Thisinhibition occurs after treatment with

SCIENCE, VOL. 207

nalidixic acid or coumermycin A1 (77)and temperature shift upwards of a gyrAtemperature-sensitive mutant (29, 30).The rapidity of the inhibition implies arole in the elongation phase of replica-tion. Gyrase could function as an activeversion of the replication "swivel" firstpostulated by Cairns (78) to relieve thepositive twisting stress generated by theunwinding of the double helix during rep-lication (79). Alternatively, or additional-ly, the maintenance of negative super-helicity by gyrase may be required forthe binding of replication proteins. Thisis exemplified in the early finding thatDNA gyrase is required for OX174 DNAreplication. Supercoiling of 4X174 repli-cative form DNA is needed for the initialnicking by cisA protein (80), and thusDNA synthesis is blocked by nalidixicacid or novobiocin (81). In the recentachievement of replication of 4X174DNA with the use of only purified pro-teins, DNA gyrase completed the requi-site set of enzymes (82). Novobiocin hasan interesting selective effect on B. sub-tilis DNA synthesis (83). Only a limitedsegment of the chromosome at the originis replicated in the presence of the drugand this allowed precise mapping of thisregion.

Gyrase is required for other metabolicprocesses involving DNA as substrate ortemplate. In their evolution, such pro-cesses may have taken advantage of thesupercoiling of DNA for recognition andactivity (22). This seems to be the casefor at least some aspects of DNA tran-scription, repair, and recombination.Treatment with gyrase inhibitors or

mutational inactivation of gyrase gener-ally results in only a partial inhibition ofoverall RNA synthesis (29); however,nalidixic acid completely blocks phageS13 transcription (84). The resolution ofthis puzzle seems to be that the depen-dence of transcription on gyrase variesfor different operons (85-90). The verysame gene may be either sensitive or re-sistant, depending on the promoter it isread from (87, 88, 90). An illuminatingexample is that of phage N4 early tran-scription, which is obliterated by nalidix-ic acid, coumermycin A1, or gyrA condi-tional lethal mutations (86, 91). The puri-fied N4 RNA polymerase transcribes on-ly single-stranded N4 DNA in vitro.However, in the cell it is likely that su-percoiling of duplex N4 DNA facilitateslocal denaturation of the duplex andbinding of the polymerase to the appro-priate strand. A similar explanation mayapply to other instances of gyrase effectson transcription, and the assistanceneeded could vary depending on the easeof helix unwinding at the promoter site.29 FEBRUARY 1980

Consistent with this is the report that gy-rase inhibitors selectively interfere withthe initiation of transcription (90). Pro-moters are generally associated with pa-lindromic DNA sequences and theseloop out to form hairpin-like structureswhen contained in supercoiled DNAgenerated by gyrase (22).The role of gyrase in X int-promoted

recombination has been demonstratedand, indeed, led to the initial discoveryof supertwisting activity. CoumermycinA1 effectively blocks both supercoiling ofX DNA and int-promoted recombinationin vivo and in vitro, and both effects areabsent in a drug-resistant gyrB mutant(71, 92). Furthermore, the requirementfor gyrase in this recombinational pro-cess is simply to provide negative super-helicity, since the enzyme is totally dis-pensable when a supertwisted DNA sub-strate is provided (14). An involvementof gyrase in other recombination path-ways is also possible. The quintessenceof recombination is transient double-strand breaks that are inherent in the gy-rase reaction mechanism, and a logicalstart point for recombination is the inter-action of single-stranded DNA with a ho-mologous supertwisted DNA duplex(93). The properties of DNA cleavage bygyrase have also led to a suggested rolein the transposition of genetic elements,an example of nonhomologous recombi-nation (94).Whereas early experiments implied

that repair of damaged DNA was resist-ant to gyrase antagonists (77), a recentmore sensitive study based on the resto-ration of infectivity of ultraviolet-irra-diated phage X DNA has shown clear in-hibition (95). It is likely that only certainrepair pathways are dependent on gy-rase-in fact, nalidixic acid induces therecA gene product and an error-proneDNA repair pathway (96).An exciting possibility is that the new-

ly discovered catenation, uncatenation,knotting, and unknotting activities ofDNA gyrase (40, 51) are also physiologi-cally relevant. Catenanes are presentwherever there are DNA rings and havebeen suggested as intermediates in thetermination of replication and in recom-bination (48). Convincing evidence fortheir resolution into constituent mono-mers has been lacking (49), but the dis-covery of efficient uncatenation by gy-rase provides the first enzymatic mecha-nism for the process. DNA can be knot-ted during packaging into phage heads(97) and, upon infection, the knots needto be untied, presumably by a topo-isomerase such as gyrase.

While a fundamental role of gyrase invarious physiological processes is clear,

there are several complications to con-sider in interpreting physiological stud-ies. First, the gyrA gene product is notjust a part of gyrase but also of topo II'(24). It is thus important to determine theinvolvement of both gyrase subunits inany process. Second, it is likely that nali-dixic acid does not just inactivate its tar-get protein but corrupts it, converting itinto a poison (30). Phage T7 is ex-quisitely sensitive to nalidixic acid, butmutational inactivation of either thegyrA or gyrB gene product has no effecton T7 growth or DNA synthesis and,more importantly, prevents inhibition bynalidixic acid (30, 98). Therefore, resultswith drugs must be cross-checked withmutants. Third, an enzyme in vitro mightneed superhelical free energy to facilitatebinding to DNA, but in the cell some oth-er factor may enhance binding. Thesethree limitations suggest caution but, for-tunately, the convergence of differentapproaches has solidified many aspectsof gyrase function.

Comparison of Origin of Supercoiling in

Prokaryotes and Eukaryotes

With rare exceptions, circular DNAisolated from all natural sources is nega-tively supercoiled to about the same de-gree (5). Supercoils in the chromatin ofeukaryotes almost surely arise from neg-ative coiling of the DNA around histoneoctets to form nucleosome units (99) fol-lowed by enzymatic relaxation of com-pensatory positive supercoils (100, 101).Thus, as seems to be true for gyrase, theultimate motive force for supercoiling inchromatin is spontaneous wrapping ofDNA around a protein core. Even theextent of wrapping seems the same. Gy-rase and the histone octet each protectabout 140 base pairs of DNA from diges-tion by staphylococcal nuclease and anestimated one and a quarter supercoilsfrom relaxation (21, 23, 47, 99). This sim-ilarity is surprising since the 400,000-dal-ton gyrase is much larger than the110,000-dalton histone assembly.The similarities in formation of super-

coils by prokaryotes and eukaryotesseem to result not from evolutionary ho-mology but from convergence to performthe same function. A distinguishing fea-ture of histones is their basicity, andboth gyrase subunits are acidic (21, 28).The topoisomerase and the spool forDNA are separate entities in chromatinbut are united in gyrase. Gyrase acts cat-alytically and requires ATP, whereas su-percoiling in chromatin is stoichiometricwith the number of nucleosomes. Themost fundamental difference is that gy-

959

rase protects a positive supercoil, butDNA is negatively supercoiled in nucle-osomes. The free negative coils pro-duced by gyrase represent a physiologi-cally significant source of energy and areready to assist in separating the double-helical strands (102), whereas the nucle-osomal supercoils would have to befreed from the stabilizing histones beforethe energy of supercoiling could be har-nessed.

References and Notes

1. M. Gellert, K. Mizuuchi, M. H. O'Dea, H. A.Nash, Proc. Natl. Acad. Sci. U.S.A. 73, 3872(1976).

2. P. 0. Brown and N. R. Cozzarelli, Science 206,1081 (1979).

3. For earlier reviews of gyrase see D. T. Den-hardt, Nature (London) 280, 196 (1979) and (9,19, 21, and 22).

4. J. Vinograd, J. Lebowitz, R. Radloff, R. Wat-son, P. Laipis, Proc. Natl. Acad. Sci. U.S.A.53, 1104 (1965).

5. W. R. Bauer, Annu. Rev. Biophys. Bioeng. 7,287 (1978).

6. A description of the topology of supercoils isgiven by F. H. C. Crick, Proc. Natl. Acad. Sci.U.S.A. 73, 2639 (1976) and in (5).

7. Y. Hayashi and M. Hayashi, Biochemistry 10,4212 (1971); J. C. Wang, in DNA Synthesis:Present and Future, I. J. Molineux and M. Ko-hiyama, Eds. (Plenum, New York, 1978), pp.347-366.

8. J. C. Wang, J. Mol. Biol. 55, 523 (1971).9. __ and L. F. Liu, in Molecular Genetics, J.

H. Taylor, Ed. (Academic Press, New York,1979), part 3, pp. 65-88.

10. V. T. Kung and J. C. Wang, J. Biol. Chem.252, 5398 (1977).

11. J. J. Champoux and R. Dulbecco, Proc. Natl.Acad. Sci. U.S.A. 69, 143 (1972).

12. J. J. Champoux, Annu. Rev. Biochem. 47, 449(1978).

13. K. Mizuuchi and H. A. Nash, Proc. Natl.Acad. Sci. U.S.A. 73, 3524 (1976).

14. K. Mizuuchi, M. Gellert, H. A. Nash J. Mol.Biol. 121, 375 (1978).

15. M. Gellert, M. H. O'Dea, T. Itoh, J. Tomi-zawa, Proc. Natl. Acad. Sci. U.S.A. 73, 4474(1976).

16. M. J. Ryan, in Antibiotics, F. E. Hahn, Ed.(Springer, Berlin, 1978), vol. 5, pp. 214-234.

17. A. Sugino, C. L. Peebles, K. N. Kreuzer, N.R. Cozzarelli, Proc. Natl. Acad. Sci. U.S.A.74, 4767 (1977).

18. G. J. Bourguignon, M. Levitt, R. Sternglanz,Antimicrob. Agents Chemother. 4, 479 (1973).

19. C. Sumida-Yasumoto, J.-E. Ikeda, E. Benz, K.J. Marians, R. Vicuna, S. Sugrue, S. L. Zi-pursky, J. Hurwitz, Cold Spring Harbor Symp.Quant. Biol. 43, 311 (1978).

20. M. Gellert, K. Mizuuchi, M. H. O'Dea, T.Itoh, J. Tomizawa, Proc. Natl. Acad. Sci.U.S.A. 74, 4772 (1977).

21. C. L. Peebles, N. P. Higgins, K. N. Kreuzer,A. Morrison, P. 0. Brown, A. Sugino, N. R.Cozzarelli, Cold Spring Harbor Symp. Quant.Biol. 43, 41 (1978).

22. M. Gellert, K. Mizuuchi, M. H. O'Dea, H. Oh-mori, J. Tomizawa, ibid., p. 35.

23. L. F. Liu and J. C. Wang, Proc. NatI. Acad.Sci. U.S.A. 75, 2098 (1978).

24. P. 0. Brown, C. L. Peebles, N. R. Cozzarelli,ibid. 76, 6110 (1979).

25. N. P. Higgins, C. L. Peebles, A. Sugino, N. R.Cozzarelli, ibid. 75, 1773 (1978).

26. A. Sugino and N. R. Cozzarelli, J. Biol.Chem., in press.

27. K. Mizuuchi, M. H. O'Dea, M. Gellert, Proc.Natl. Acad. Sci. U.S.A. 75, 5960 (1978).

28. F. G. Hansen and K. von Meyenburg, Mol.Gen. Genet. 175, 135 (1979).

29. K. N. Kreuzer, K. McEntee, A. P. Geballe, N.R. Cozzareili, ibid. 167, 129 (1978).

30. K. N. Kreuzer and N. R. Cozzarelli, J. Bacte-riol. 140, 424 (1979).

31. A. Sugino, N. P. Higgins, N. R. Cozzarelli, inpreparation.

32. L. F. Liu and J. C. Wang, unpublished data.33. A. Morrison, N. P. Higgins, N. R. Cozzarelli,

J. Biol. Chem., in press.34. As emphasized (21), the more widely used term

of nicking-closing activity is too restrictivesince it implies that the intermediate is neces-sarily nicked and not broken across bothstrands.

35. A. Kornberg, DNA Replication (Freeman, SanFrancisco, 1980).

36. A. Sugino, N. P. Higgins, P. 0. Brown, C. L.Peebles, N. R. Cozzarelli, Proc. Natl. Acad.Sci. U.S.A. 75, 4838 (1978).

37. The effects of nalidixic and oxolinic acids arequalitatively identical, as are those of novobio-cin and coumermycin A1. For brevity, I usuallyrefer to only one member of each pair of re-lated inhibitors.

38. A. Morrison and N. R. CozzareHli, Cell 17, 175(1979).

39. See N. P. Higgins and N. R. Cozzarelli [inMethods Enzymol. 68, 50 (1979)] for a com-parison of DNA ligases with other DNA join-ing enzymes.

40. K. N. Kreuzer and N. R. Cozzarelli, unpub-lished data.

41. A. Sugino and K. Bott, unpublished data.42. Recent genetic evidence (K. Bott, personal

communication) suggests that subunit A of B.subtilis gyrase can substitute in E. coli for theresident subunit A.

43. S. Gerrard, A. Morrison, N. R. Cozzarelli, un-published data.

44. N. P. Higgins, A. Morrison, N. R. Cozzarelli,unpublished observations.

45. M. Snyder and K. Drlica, J. Mol. Biol. 131, 287(1979).

46. A. Worcel and E. Burgi, ibid. 71, 127 (1972).47. L. F. Liu and J. C. Wang, Cell 15, 979 (1978).48. H. Kasamatsu and J. Vinograd, Annu. Rev.

Biochem. 43, 695 (1974).49. J. Tomizawa, in DNA Synthesis: Present and

Future, I. J. Molineux and M. Kohiyama, Eds.(Plenum, New York, 1978), pp. 797-826.

50. L. F. Liu, C.-C. Liu, B. M. Alberts, Nature(London) 281, 456 (1979).

51. __, unpublished data.52. T. L. Hill, Proc. Natl. Acad. Sci. U.S.A. 64,

267 (1969).53. Y. Kaziro, Biochim. Biophys. Acta 505, 95

(1978).54. E. W. Taylor, CRC Crit. Rev. Biochem. 6, 103

(1979).55. P. D. Boyer, in Bioenergetics of Membranes

(Addison-Wesley, Reading, Mass., in press).56. F. Lipmann, Adv. Enzymol. 1, 99 (1941).57. H. M. Kalckar, Chem. Rev. 28, 71 (1941).58. R. P. Cunningham, T. Shibata, C. Das Gupta,

C. M. Radding, Nature (London) 281, 191(1979); T. Shabata, R. P. Cunningham, C. DasGupta, C. M. Radding, Proc. Natl. Acad. Sci.U.S.A. 76, 5100 (1979); K. M. McEntee, G.Weinstock, I. R. Lehman, unpublished data.

59. An interesting exception to this generalizationis the model proposed by P. Forterre (in prep-aration).

60. The topology of this transition is discussed byF. B. Fuller, Proc. Natl. Acad. Sci. U.S.A. 75,3557 (1978).

61. R. J. Otter and N. R. Cozzarelli, unpublisheddata.

62. G. L. Stetler, G. J. King, W. M. Huang, Proc.Natl. Acad. Sci. U.S.A. 76, 3737 (1979).

63. J. P. Baird, G. J. Bourguignon, R. Sternglanz,J. Virol. 9, 17 (1972).

,64. D. McCarthy, J. Mol. Biol. 127, 265 (1979).65. While this property sets the T4 topoisomerase

apart from gyrase, it should be noted that ATPgreatly stimulates the breakage-and-reunionactivity of gyrase.

66. D. McCarthy, C. Minner, H. Bernstein, C.Bernstein,J. Mol. Biol. 106, 963 (1976); W. M.Huang, Cold Spring Harbor Symp. Quant.Biol. 43, 495 (1978).

67. D. G. Attardi, G. Martini, E. Mattoccia, G. P.Tocchini-Valentini, Proc. Natl. Acad. Sci.U.S.A. 73, 554 (1976).

68. G. P. Tocchini-Valentini and M. I. Baldi, per-sonal communication.

69. T. Hsieh and D. M. Brutlag, unpublished data.70. P. T. Englund, personal communication.71. Y. Kikuchi and H. Nash, Cold Spring Harbor

Symp. Quant. Biol. 43, 1099 (1978).72. K. Drlica and M. Snyder, J. Mol. Biol. 120, 145

(1978).73. M. Waring, in Antibiotics: Mechanism of Ac-

tion ofAntimicrobial and Antitumor Agents, J.W. Corcoran and F. E. Hahn, Eds. (Springer-Verlag, New York, 1975), vol. 3, pp. 141-165.

74. G. L. McHugh and M. N. Swartz, Antimicrob.Agents Chemother. 12, 425 (1977).

75. D. E. Taylor and J. G. Levine, Mol. Gen. Gen-et. 174, 127 (1979).

76. J. Rouviere-Yaniv, M. Yaniv, J. E. Germond,Cell 17, 265 (1979).

77. Reviewed in N. R. Cozzarelli, Annu. Rev. Bio-chem. 46, 641(1977).

78. J. Cairns, J. Mol. Biol. 6, 208 (1963).79. Although both introduction of negative super-

coils and removal of positive supercoils reducethe linking number by changing writhing, theycan have very different consequences. Just adiminution without complete removal of posi-tive supertwists would suffice for a swivel; butinstead of providing energy for binding en-zymes that separate strands, the remainingpositive supertwists would hinder their com-plexation.

80. K. J. Marians, J. Ikeda, S. Schlagman, J. Hur-witz, Proc. Natl. Acad. Sci. U.S.A. 74, 1965(1977).

81. C. Sumida-Yasumoto, A. Yudelevich, J. Hur-witz, ibid. 73, 1887 (1976).

82. J. Shlomai and A. Kornberg, personal commu-nication.

83. N. Ogasawara, M. Seiki, H. Yoshikawa, Na-ture (London) 281, 702 (1979).

84. A Puga and I. Tessman, J. Mol. Biol. 75, 99(1973).

85. H. Shuman and M. Schwartz, Biochem.Biophys. Res. Commun. 64, 204 (1975).

86. S. C. Falco, R. Zivin, L. B. Rothman-Denes,Proc. Natl. Acad. Sci. U.S.A. 75, 3220 (1978).

87. C. L. Smith, K. Kubo, F. Imamoto, Nature(London) 275, 420 (1978).

88. B. Sanzey, J. Bacteriol. 138, 40 (1979).89. H.-L. Yang, K. Heller, M. Gellert, G. Zubay,

Proc. Natl. Acad. Sci. U.S.A. 76, 3304 (1979).90. M. Kubo, Y. Kano, H. Nakamura, A. Nagata,

F. Imamoto, Gene 7, 153 (1979).91. C. Malone and L. B. Rothman-Denes, unpub-

lished data.92. It has been found that supertwisting is dis-

pensable for the in vitro reaction at low (non-physiological) salt concentrations [T. J. Pol-lock and K. Abremski, J. Mol. Biol. 131, 651(1979)].

93. Reviewed in C. M. Radding, Annu. Rev. Bio-chem. 47, 847 (1978).

94. J. Shapiro, Proc. Natl. Acad. Sci. U.S.A. 76,1933 (1979).

95. J. B. Hays and S. Boehmer, ibid. 75, 4125(1978).

96. E. Witkin, Bacteriol. Rev. 40, 869 (1976).97. J. C. Wang, L. L. Liu, R. Calendar, unpub-

lished data.98. J. J. Donegan and R. Sternglanz, unpublished

data.99. R. Kornberg, Annu. Rev. Biochem. 46, 931

(1977); G. Felsenfeld, Nature (London) 271,115 (1978).

100. J. E. Germond, B. Hirt, P. Oudet, M. Gross-Bellard, P. Chambon, Proc. Natl. Acad. Sci.U.S.A. 72, 1843 (1975).

101. R. A. Lasky, B. M. Honda, A. D. Mills, J. T.Finch, Nature (London) 275, 416 (1978).

102. B. M. Alberts and R. Sternglanz, ibid. 269, 655(1977).

103. My co-workers in the study of DNA gyrasehave been P. 0. Brown, S. Gerrard, N. P. Hig-gins, K. N. Kreuzer, A. Morrison, R. Otter, C.L. Peebles, and A. Sugino. I am indebted to A.Morrison, K. N. Kreuzer, and P. 0. Brown fortheir aid in preparing this article.

SCIENCE, VOL. 207960